The trees that could never have children: ninety-six days with nowhere to wait
Oak, pine and hazel drop their seed in autumn, and that seed will only sprout in spring. The Long Watch asks both questions in the same instant, and autumn is never spring, so every acorn those trees have cast was discarded on the ground where it landed. Three of the world’s twenty-five plants have been sterile since the day seasons shipped.
A clean bill of health over sixteen dead plants
We found it on the seventeenth of July, re-checking the plant systems after the merge that put every climate into a single world. The check that watches canopy seeding printed its active banner and signed off that stability was confirmed. What it had measured was sixteen plants, every one of them already in the final dead state, with not one seed added across the whole run. Both plant checks had been read and accepted before, over the same dead ground.
Nothing in the bookkeeping was wrong, which is why the failure had never made a sound — the shape of green we had already learned to distrust. Every counter that asked whether dispersal had run came back yes, and every one of them was telling the truth. The parents did disperse. The seed went nowhere.
Autumn is never spring
Seasons arrived at the end of May, and with them two new facts per species: the season a plant releases its seed, and the season or seasons in which a seed of its kind may germinate. The three canopy trees were given the real biology. Acorns fall in autumn. Acorns come up the following spring.
The step that scatters seed asks both of those questions at once. It walks the mature plants, casts their seed out across the ground, and decides on the spot whether each landing becomes a seedling. The moment it tests for germination is the moment the seed was cast. For a species that drops in autumn and sprouts in spring, that test can only ever fail. Where a seed lands is a careful piece of work, and none of it mattered, because the landing was refused on arrival.
A year here runs a hundred and ninety-two days, four seasons of forty-eight. Between an autumn seed-fall and the spring window sit ninety-six of them — half a year, with nowhere in the simulation for a seed to spend it. There was no seed bank. There never had been one.
Not sometimes, and not in certain worlds. Three canopy species could never set a single seedling, at any moment, in any world, forever.
Lifting the gate, and counting all twenty-five
A defect that total invites disbelief, so we ran the control. In the dispersal scenario a hundred and fifty-two mature parents stood ready to seed and produced nothing whatever. We lifted the germination test, left everything else alone, and replayed it: a hundred and twenty-one seedlings came up, and the plant count went from a hundred and fifty-seven to two hundred and seventy-eight. The trees had been fertile the entire time, and gated out of breeding.
“Exactly three species” is a claim worth more than the spot-check that suggests it, so we enumerated the roster rather than sampling it — a roster that had grown on description alone. Of twenty-five plants, three carry the fatal pairing. The other twenty-two either overlap their two seasons or ignore the calendar altogether, and no repair we make will change a thing about them. A fourth canopy tree, the swamp bloodwood written for the wetlands in July, was authored with open seasons and reproduces without trouble.
The scenario the canopy check runs against had been cloned from the oak when it was first written, sterile pairing and all. The check was infertile for precisely the reason the world was.
The plan said both things, one sentence apart
The document that introduced the seasonal settings described an acorn that overwinters and wakes in spring. A sentence later, specifying the mechanism, it said an out-of-season seedling is filtered out at the landing gate. Those two sentences cannot both be built. The second is the one that got written, and it was reviewed and approved exactly as written.
The intent required persistence. The ratified mechanism guaranteed discard. Both were approved.
The same day’s re-check turned up a second seasonal defect beside this one, in the arithmetic of growth, where a stray wobble had been cancelling winter for every plant in the world; that one was repaired the same afternoon. This one we left alone deliberately. Turning “falls in autumn, comes up in spring” into something the game can actually do is a decision about how this world works, not a patch.
What a seed would have to be
Two cheap escapes were available. Make the three species ignore the calendar, or slide their seasons until dropping and sprouting overlap. Either takes an afternoon, and either throws away the thing worth keeping. The acorn’s biology is right; the machinery has to grow up to it.
So we wrote a design for what a seed actually is. A seed is not a seedling. A dispersed seed goes into a seed bank and waits there from the tick it lands until its window opens. While it waits it does nothing at all: it never grows, it is never drawn, it feeds no animal, and it gives nothing back to the ground. It carries two facts — where it fell, and when.
Keeping banked seed out of the living-plant list is a deliberate exclusion rather than an optimisation. Anything in that list can be grazed, drawn, counted, and blessed, and a buried acorn should be none of those. No forager should eat what it could not have found, and no player should be able to lay a blessing on ground that reads as bare.
The bank also has to be written down. Almost everything in a Long Watch world is rebuilt on load rather than stored, but a seed lying in the soil cannot be re-derived from the world seed and the clock, so it becomes one of the few things a save has to carry. Put a world down in autumn, come back to it in winter, and the year’s crop is still there, waiting on a spring that will now actually arrive.
Dormancy is declared per species, never inferred from two seasons happening not to overlap. Inference is how we arrived here. A plant written a year from now whose windows miss each other should fail loudly rather than acquire a seed bank nobody designed for it.
Two losses come with the design, both intended. Not every banked seed is still viable when its window opens, and animals eat seed while it waits. The second ships switched off, because a forager reaching into the bank would couple what the camera happens to be simulating to what the world writes down, and that is a boundary we have already had to defend. Neither loss is a population controller; nothing watches the bank and tops it up. A species whose seed is all eaten thins and retreats to the ground that suits it, which is the ecology working rather than failing.
The forest that has not aged
The plan is written and waiting on approval. What that day actually shipped were guards: both plant checks now fail outright when nothing is alive in the scenario beneath them. No dormancy code exists in the game yet, and the forest still cannot renew itself.
The design is candid about one thing it will not be able to prove. The determinism checks covering canopy dispersal watch about a day of in-game time against a seed that has to survive ninety-six days of them, so as they stand they can never witness dormancy from fall to sprout. We would rather write that down than shrink the wait until a green result becomes purchasable.
What this phase is aiming at is a forest that has aged: a late world whose regions have visibly become themselves, a wood thicker in one valley than the next because of what happened there. Three canopy species that cannot set a seedling can only thin out from whatever world-generation put down. An oak here takes in-game years to finish dying, standing, then fallen, then litter, giving back everything it has to the soil beneath it. Nothing has ever come up out of that ground.
